EP3791359A1

EP3791359A1 - Correlation of thermal satellite image data for generating thermal maps at high spatial resolution

Info

Publication number: EP3791359A1
Application number: EP19724144.1A
Authority: EP
Inventors: Max GULDE
Original assignee: Fraunhofer Gesellschaft zur Forderung der Angewandten Forschung eV
Current assignee: Fraunhofer Gesellschaft zur Forderung der Angewandten Forschung eV
Priority date: 2018-05-09
Filing date: 2019-05-08
Publication date: 2021-03-17
Anticipated expiration: 2039-05-08
Also published as: ES2927925T3; JP7334879B2; DE102018207265A1; WO2019215210A1; JP2021523450A; CN112106106A; CN112106106B; US20210319590A1; EP3791359B1

Abstract

The invention relates to a method and a device for producing a thermal map of an area, wherein the thermal map is generated by a combination of two thermal images of different properties, and both thermal images comprise pixels associated with the area and have been recorded by satellites. The two thermal images are recorded at different times using different recording devices. Furthermore, a radiometric precision of the first thermal image is higher than that of the second thermal image, and a spatial resolution of the second thermal image is higher than that of the first thermal image. The two thermal images are used to determine a measurement value offset of a first pixel group belonging to the second thermal image and spatially associated with the area, and then corrected absolute measurement values of the pixels belonging to the second thermal image and spatially associated with the area are determined. A precise thermal map of the area is then created on this basis. The invention also relates to a method for determining a time series of thermal maps and a computer programme product for carrying out the method described.

Description

Correlation of thermal satellite image data to generate spatially high resolution heat maps

The present invention relates to the determination of spatially high resolution thermal maps using satellite image thermal data.

An as accurate as possible determined land surface temperature (LST, Land Surface Temperature) is an important part of many different applications, for example in the detection of forest fires, the measurement of local maxima in the urban temperature distribution (UHI, Urban Heat Is land), in the determination of vegetation indices in agriculture or modeling the local and global climate. The large-area and accurate measurement of the surface temperature of the land has become possible only with the use of satellites.

The temperature is not measured directly, for example, but derived from the radiance at the detector aboard the satellite. The radiant Te Rx is the radiation (energy) emitted by a surface in a given time in a given spectral range in a given direction and is given in watts per square meter steradian.

The detector itself records a gray level image in which the brightness of a single pixel can be assigned to a measured radiance.

The assignment of gray values DN to radiance R occurs, for example, by a linear equation of the form

R = a DN + b

, wherein the constants a and b are determined by a calibration of the detector ermit.

Via the Planck radiation law, at a known wavelength l di, it is possible to directly infer the temperature T via the following relationship:

k = 2 hc ²

k ₂ = hc / a _B

Here h is Planck's constant, c the speed of light in vacuum and s _B the Boltzmann constant.

The measured output from which the temperature can be derived may depend on the type of detector. In so-called micro bolometers, for example, either a temperature-dependent electrical shear resistance is measured or the temperature change of the absorber directly through a thermometer. In the first case, the temperature is determined by a previously determined functional dependence between resistance and temperature.

Mercury Cadmium Telluride (English MCT) detectors are semiconductor devices in which the amount of energy trapped per pixel correlates linearly with its brightness in a good approximation. With a known field of view of the pixel and exposure time can thus be concluded on the radiance which is related to temperature as described above.

Another detector type is the quantum well infrared photodetector (English QWIP), in which also charge carriers are released at Infra red (IR) light incidence, which then generate a measurable current.

There are other types of detectors, but they usually have in common that the temperature is determined as a derived quantity by changing easily measurable physical parameters (resistance, voltage, etc.).

A general challenge in the measurement of LST by infrared cameras from space is the achievement of a high absolute radiometric accuracy, ie accuracy of the measured radiation temperature of the earth's surface. The measured temperature should reflect the actual radiation temperature of the measuring range on the average as accurately as possible. This results in difficulties, inter alia, by temperature fluctuations of the detector, the passage of the IR radiation to be measured through the atmosphere, the Weil length and viewing angle-dependent Emis sion level of the earth's surface, different surface types (such as roads, fields, forests, buildings, etc.) with different emissivities in the measuring range, the reflected solar radiation and the physical interpretation of the results. Thus, a radiation temperature is usually measured and an attempt is then made to conclude the surface temperature by a surface model.

Many of the aforementioned applications simultaneously require both high temporal and high spatial resolution to record, for example, the day cycle of UHIs or evapotranspiration. This requirement can not be fulfilled by currently existing satellite platforms planned for the next few years.

Geostationary satellites such as GOES or Meteosat have very high temporal repetition rates (5 minutes for MSG SEVIRI in rapid scanning mode) and high radiometric precision, but limited spatial resolution with projected pixel sizes of several square kilometers. Spatially high-resolution platforms such as Landsat-7 or 8 (ETM +, TI RS) and Terra (ASTER) offer spatial resolutions of around 60 - 120 m, but they only allow global coverage with a repetition rate of about two weeks.

In general, high-quality calibration methods for precise measurement of the surface temperature on board the satellite are very expensive and so far only to be realized within large satellites. These platforms have the disadvantage that at high spatial resolution (eg Landsat 7 and 8) in low orbits very long time passes until the same area can be observed again be switched and use of a constellation of a variety of such systems an extreme financial cost meant. Remote platforms (Meteosat, GEOS) suffer from the problem of poor spatial resolution as described above.

There are therefore several approaches to generate radiometrically precise thermal data with high spatial and optionally also high temporal resolution. The approaches all have in common to combine existing precise thermal satellite data of large platforms such as Meteosat and Landsat, which have elaborate calibration methods on board, with spatial higher-resolution auxiliary data. Such a downsampling of thermal data can be used to obtain spatially higher resolution thermal images and to use them, for example, in the context of a time-resolved high-resolution time series.

There are currently two types of procedures to achieve this spatial resolution increase. Thermal Sharpening (TSP) uses high-resolution physical environment models to predict temperatures in a rectangular subpixel network. These can be, for example, maps with exact information on emission coefficients or the local Al bedo. It is assumed that the recorded, coarse Tempe raturpixel can be expressed as a linear combination of its Subpixelkonstituenten. These are divided into classes depending on the parameter set (eg "forest", "meadow", "street", "building") and correlated, for example, via multilinear regression or sequential Monte Carlo methods. The emission coefficient e determines the relationship between the measured radiation temperature T _s and the kinetic surface temperature T _{0 of} a body via the following relationship:

Ts ^e T ₀

Remote Sensing methods typically measure the product of temperature and emission coefficient. Since both variables can not be determined independently in a band at the same time, the problem is then underdetermined. Accordingly, further information is needed. There are several ways to determine the emission coefficient.

On the one hand, a map of the observed area with detailed information on the type of landscape (CORINE is one such map for Europe, which is recreated approximately every two years). If the type of landscape is known, so-called spectral libraries (Gillespie et al., 1998) can be used to read and allocate the associated emission coefficients (or closely related reflection coefficients).

Alternatively, when using multiple spectral bands (at least 3), a separation of emission coefficients and temperature can take place. For vegetation-covered areas is also often above the normalized

Difference Vegetation Index (NDVI) on the emission coefficients (Valor & Caselles 1996).

The application of TSP can lead to relatively precise results with a moderate increase in resolution (about factor 10) and good data situation, with average LST errors (RMSE) of about 2-3 K. The results are all the more accurate, the fewer classes and subcomponents are chosen. With actual application to Meteosat data, the factor of the increase in resolution (downsampling) depending on the application is about 40 - 100 and there are many classes within a coarse pixel (pixel size of SEVIRI minimal 9.6 km ² , about Central Europe about 15 km ² ) , The mean error here is at best about 5 K, but individual outliers may well differ by 20 K and more.

Alternatively, spectral segregation (TUM, Temperature UnMixing) attempts to infer irregularly shaped subpixel components. In the case of sem process, for example, several thermal bands are analyzed in parallel, in order to achieve an increase in resolution of the output data. Alternatively, temporal image sequences of the same region, different angles of view or different resolutions in further, synchronously recorded thermal bands can be used to improve the spatial resolution of a region. However, there are few satellite sensors that record regions from multiple angles, the emission coefficient needed is difficult to determine when using multiple bands, and time sequences are difficult to parametrize if the surface parameters are time-variable.

However, lower-priced small satellites with high spatial resolution are not able to capture images with high radiometric precision. When using smaller satellite platforms, for example pico- or nanosatellites, approaches for the generation of spatially high-resolution thermal data of the half mostly based on the calibration of spatially high-resolution IR images together with reference objects known temperature. By recognizing the reference objects within the IR image, this can then be calibrated accordingly. Alternatively, a second sensor is used which provides precise temperature information of an image section. The use of cooled detectors on small satellites is also made, but is usually not sufficient to determine the temperature to less than a few Kelvin.

There is therefore a need for simple, cost-effective solutions for the creation of spatially high-resolution thermal images with high radiometric precision, as well as for the generation of temporally high-resolution time series consisting of various spatially high-resolution thermal images of high radiometric precision.

The present patent application solves this problem according to the independent claims. Advantageous further developments are also described in the dependent claims and this description.

The proposed solution consists in the use of small satellites to create spatially high-resolution heat maps of the earth's surface at the same time high radiometric precision. The presented principle is based on the combination of 1) recorded with satellite ther maldaten high absolute accuracy, but low spatial Auflö solution with 2) temporally and spatially ko registered, ie temporally synchronous and from the same area, recorded thermal data of a second satellite platform shape with lesser Measuring accuracy but significantly higher spatial resolution.

In this case, measured values are, for example, the gray values of a photosensor, which can be associated with a physical radiance (and thus radiation temperature), the electrical resistance or an electrical voltage which can be directly assigned to a temperature or other physical measured variables used for non-contact determination of the temperature ,

The method described here and the device combine the advantages of both measurements and thus results both in a high tempera ture precision as well as in a high spatial resolution. At the same time, uncertainties arising from the usual methods for the spatial disaggregation of satellite-based LST can be avoided, at least in large parts. Furthermore, a method for creating a time series of spatially high-resolution thermal maps and a computer program product for carrying out the described method will be described.

The method described is designed to determine the thermal map of a region, preferably to determine a spatially highly resolved th thermal map with a high measurement accuracy, such as a high temperature accuracy.

The method comprises receiving a first thermal image and receiving a second thermal image, wherein the thermal images are provided by different platforms and therefore have different properties.

A thermal image is a graphical representation of a physical unit over a region and can represent various physical units. set, for example, a radiance (W / m ² sr) or a temperature (K).

The first thermal image taken by a receiving device of a first satellite has a high radiometric precision but a low spatial resolution. The first thermal image shows the recorded thermal radiation of a first landscape comprising the region, wherein the first thermal image comprises pixels spatially associated with the region and the first thermal image pixel-wise each a measured first temperature value, radiance, or other radiometry assigns to the measured variable.

The second thermal image taken by a cradle of a second satellite or, alternatively, a drone, a balloon or other manned or unmanned aerial vehicle, has a lower radiometric precision but a higher spatial resolution, compared to the first thermal image. The second thermal image shows the recorded thermal radiation of a second landscape comprising the area, the second thermal image comprising pixels spatially associated with the area and the second thermal image associating the area pixel by pixel with a measured second temperature value.

The "area" for which the heat map is to be determined may well be a small part of the landscape, for example a single building or field, or even a larger part of the landscape shown on the heat pictures, such as a village , a district or an entire city. The "area" can thus extend over various pixels of at least the second thermal image.

Different pixels can be assigned to the individual pixels of the two thermal images. In addition, the spatial resolution of the second thermal image is higher than the spatial resolution of the first thermal image, so that in each case a plurality of pixels of the second thermal image are spatially associated with one pixel of the first thermal image.

The first and second thermal images continued to be temporally synchronous or less than a predetermined time delay Barrier of preferably a few minutes, for example, taken 10 minutes. The temporal tolerance is given by the characteristic period of time with which the temperature or the radiation properties in the area change significantly. This period of time may vary depending on the conditions, such as the weather conditions, as well as the intended use case.

The high radiometric precision and thus temperature accuracy or, more generally, measurement accuracy of the first thermal image results beispielswei se from the receiving device of the first satellite, which is preferably a large satellite, such as a weather satellite, and has sophisticated calibration technology, for example, using specially tempered black bodies or sensors. Satellites that can capture images with high accuracy, such as temperature accuracy, are known in the art, however, the necessary Kalib ration technology is very costly and too large for small satellites such as pico, nano, or microsatellite. The first satellite can be a geo stationary satellite, which further allows a high temporal repetition rate of the recorded thermal images. However, geostationary weather satellites are not able to take high-resolution images due to their distance from Earth. Thus, the spatial resolution of the first thermal image is in the range of one to several square kilometers per pixel.

Alternatively, the first satellite may be a large satellite in low earth orbit, such as known by the Landsat satellites. Such a satellite, which is also configured via calibration technology for the acquisition of thermal images with high radiometric precision, a significantly better spatial resolution compared to a geo stationary satellite is possible. However, the spatial resolution for many applications is still in need of improvement. Furthermore, such large satellite in a low Earth orbit the same Teilaus cut the earth's surface record only at a distance of several days to several weeks, making a recording at a desired time point or a time series of several shots or only with a very low temporal resolution or by Combination of several such Satellite is possible. However, since large satellites are very costly, it is financially very expensive to use the necessary for a better time accuracy variety of large satellites in a low Earth orbit (with a distance below 2000 km to the earth).

The second thermal image can be recorded by a small satellite orbiting closer to the earth, well below 1000 kg or below 500 kg, such as a pico-satellite up to about 1 kg, a nanosatellite up to about 10-15 kg or a microsatellite up to about 100 kg , Due to the proximity to the earth, a higher spatial resolution can be made possible. Such satellites orbit in a lower Earth orbit with a distance of typically between 300 and 1000 km distance to the earth's surface, for example between 500 and 800 km distance to the earth's surface. In general, smaller and shorter life satellites orbit closer to Earth than larger satellites and / or longer lifetime satellites. However, in the case of the smaller satellites, which are several orders of magnitude cheaper (as of 2018, the cost of a geosatellite of around € 1000 million, while a nanosatellite costs approximately between € 1 and € 10 million) has the technical potential and the space for such as calibration technology on board, so that the thermal images measured by a small satellite may differ significantly from the actual measurable variable, such as temperatures or radiances. Deviations can occur, for example, due to temperature fluctuations of the infrared sensor used for recording. Since small satellites circle significantly closer to the earth than weather satellites, for example, the thermal images can be recorded by a small satellite with a much higher resolution of edge lengths of the individual pixels below 100 m, preferably below 50 m, between 30 m and 50 m or below 30 m , When post-processing the recorded images, for example, by using or overlaying several pictures taken even higher spatial resolution, example, less than 20 m or less than 10 m possible. The second satellite, with the recording device of which the second thermal image was taken, is hereby generally in a low earth orbit, in particular in a lower earth orbit than the first satellite, whereby only a lower repetition frequency, of at least several hours, can be achieved by a single satellite or days or, for example, 2 to 4 weeks at a relatively high resolution of, for example, about 100 m, for the recordings of an area is possible, please include. The repetition frequency can be considered as a function of the resolution. When spatially high resolution images are taken, the limited detector size necessarily captures a small area. Accordingly, it takes a relatively long time to cover this area again (unless the orbit has been set to cover a certain area more frequently at the expense of the total coverage of the earth's surface). Thus, the repeat frequency is a function of the field of view of the detector and thus can be directly coupled to the resolution.

The method for determining the thermal map further comprises determining a measurement offset, such as a temperature offset, of a first spatially associated pixel group of the second thermal image, wherein the first pixel group comprises a plurality of pixels. As described above, the absolute measurement accuracy, example, temperature accuracy of the second thermal image is low and can, if the measured variable, for example. To a temperature to several degrees, for example, about 2 K, or, more often, about 5 K or even about 10-20 K, deviate from reality. However, it can be assumed that the relative measured values, for example relative temperature values, of the second thermal image have sufficient accuracy, in particular the relative measured values of adjacent pixels or of pixels which are in the vicinity of one another, for example within a few pixels of one another.

It is therefore possible to determine a measured variable offset for the first pixel group by comparing a mean value of the measured value, for example the average temperature value, for the first pixel group with a value of the same measured variable of the reference pixels of the first thermal image. In this case, for example, it is also possible for adjacent pixels both with regard to the first thermal image (ie adjacent pixels of the at least one first pixel) and with respect to the second thermal image (ie adjacent pixels of the first pixel group) to be taken into account in the determination of the measured variable offset. The "coarser" reference pixels of the first thermal image can in this case be expressed as a linear combination or weighted sum of the first pixel group. The pixels of the first pixel group are taken into account in the calculation of the mean value with the weighting of their area fraction. So with also different sized pixels can be considered. The measured values can either be scalar measured quantities for each pixel, or it is possible that several values are recorded simultaneously, so that the individual measured values are then in vector form.

The method further comprises the step of determining corrected absolute measurement values of the pixels of the first pixel group assigned to the region based on the second measured measured values of the pixels of the first pixel group and the measured value offset.

The determination of the corrected absolute measured values for the pixels of the first pixel group is preferably carried out by adding the second measured measured values and the previously determined measured-value offset.

The method also includes a step of determining or generating the high spatial resolution thermal map with high radiometric precision based on the corrected absolute measurements.

By way of example, the measured variable is a temperature, measured, for example. in Kelvin. In this case, the temperature offset between the two thermal images is then determined. However, it may also be a beam size, measured e.g. in watts per square meter and per steroid, to provide electrical resistance e.g. measured in ohms or measured by a voltage e.g. in volts.

Thus, by combining two recorded thermal images of different properties, a significantly improved thermal map of the area can be created.

By the method, in particular, for example by means of infrared, the radiation temperature of the land surface is determined. Depending on the application, in addition, an accurate determination of the surface temperature of the surface given by the given heat map radiation temperature below With the aid of an emissivity of the respective surface, depending on the type of the respective surface.

It should be noted that first specific intensities or beam densities are measured by infrared sensors, which are integrated to compare the GE measured radiances of the first and second thermal image over the jewei as possible congruent surfaces. From the measured values, a radiation temperature can then be determined which can be averaged for the respective surfaces.

For this purpose, a pixel is illuminated, for example, with a certain amount of energy in the IR and thus has a certain brightness. By the well-known Pi xelantwort (eg measured in the laboratory before the start), which is usually almost linear in the area to be considered, can now be concluded by the pixel brightness on physical units, for example by means of a compensation function or a conversion table. In this case, this is the spectral radiance (W / (m ² sr pm)). The (radiation) temperature then results directly from Planck's law of radiation. The surface temperature can then be determined from this and an emission coefficient.

Due to the high spatial resolution and high radiometric precision of the thermal images generated by the described method, the determination of the surface temperature is possible with significantly reduced effort, compared with those known in the prior art method.

Preferably, the first pixel group of the second thermal image is spatially associated with exactly one pixel of the first thermal image. In this case, the measured value offset of the first pixel group of the second thermal image is preferably determined by a possibly weighted total relative measured values of the first pixel group of the second thermal image compared to the first measured measured value of the pixel of the first thermal image spatially assigned to the first pixel group of the second thermal image ,

This calculation is made on the assumption that the value of the first pixel of the first thermal image measured with high measurement accuracy an average of the measured values of the pixels of the first pixel group of the second thermal image is known. With this, the measured value offset DG can be eliminated. This is given by the difference between the mean values of, on the one hand, measured value of the first pixel of the first thermal image and, on the other hand, the measured measured values of the pixels of the first pixel group of the second thermal image.

It applies

T _A _(. X, y,) is the recorded measured size of the first thermal image of the pixel (x, y) and T _B (i, j) of the measurement of the second thermal image in all pixels (i, j) located within the first pixel group where w (j, j) represents the respective area proportion of the pixel (i) at the total area spatially associated with the first pixel. Thus, w (j, j) = 1 if the pixel (i, /) is completely contained in the total area of the first pixel.

If the temperature is chosen as the measured value, this linear approach applies, to a good approximation, to a mean temperature range on the earth of around 300 K, as normally found on a single thermal image, with a difference between the maximum and minimum Temperature of a few tens of Kelvin, for example 20 K or 30 K.

If, instead, the radiances are selected as the measured variable, the mathematical formulation corresponds to the law of conservation of energy and it applies exactly.

It is also preferably possible that after calculating the corrected absolute temperature values of the pixels of the first pixel group, the method includes optionally reworking the pixels of the first pixel group using land surface models. Such land surface models can, for example, using emission coefficients, which are associated with the respective surfaces, from corrected absolute temperature values, which, for example, include a radiation pattern. specify temperature, precise surface temperatures are calculated.

The present application further comprises a device for determining the thermal map of a region, the device comprising at least one receiving unit and a determining unit.

The receiving unit is in this case configured to receive a first thermal image of a first landscape comprising the area, wherein the first thermal image was taken by a recording device of a first satellite, wherein the first thermal image comprises pixels spatially associated with the area and the thermal image Area assigns a first measured value pixel by pixel.

The at least one receiving unit is further configured to receive a second thermal image of a second landscape comprising the region, the second thermal image being captured by a receiving device of a second satellite, the second thermal image comprising pixels spatially associated with the region second heat mebild the area pixelwise a second recorded reading points.

Between the recording of the first thermal image and the recording of the two th thermal image in this case there is only a small time offset below a specified barrier of preferably a few minutes, example, less than 30 minutes or less than 10 minutes recorded.

Furthermore, as already described with respect to the method, a ra diometric precision of the first thermal image is higher than a radiometric precision of the second thermal image, and a spatial resolution of the two th thermal image is higher than a spatial resolution of the first heat image.

The determining unit of the device for determining the thermal map of the area is configured to determine a measured value offset of a first, spatially assigned, pixel group of the second thermal image, wherein the first pixel group comprises a multiplicity of pixels a sum or linear combination of relative measured values of the pixels of the first pixel group with respect to the first recorded measured value of the at least one first pixel of the first thermal image, wherein the at least one first pixel of the first thermal image of the first pixel group of the second heat mebildes is at least partially spatially associated.

If the measured values are temperature values, the determination of a temperature offset can be carried out, for example, using formulas (1) and (2).

It should be noted here that, upon detection of radiation data, the calculation of the radiation temperature can be made either before or after determining the offsets. It is thus possible, as described in formulas (1) and (2), that the sensor data are converted into temperature values and then a temperature offset is determined by comparing the temperature values of the first and second thermal images. Alternatively, however, it is also possible for a measured value offset to be based on another measured or derived variable and the conversion into temperature values to take place only afterwards.

The determination unit is further configured to determine corrected absolute measurements of the pixels of the first pixel group based on the second acquired measurements of the pixels of the first pixel group and the measurement offset.

Below are optional, advantageous embodiments of the procedural ren and the device for determining the heat map described.

Preferably, both the first and second thermal images are infrared images respectively taken by infrared sensors of the first and second satellites, respectively.

For example, for normal LSTs in the range of large-area temperatures occurring on Earth, eg -40 to 60 ° C, infrared waves in the range of 8 - 14 pm, often even only 10.8 and 12 pm are used. For forest fires, the short-wave infrared range is more interesting, for example between 3 and 5 pm. Most weather satellites therefore have multiple bands that cover the area Cover 0.5 - 13 mih. The recorded measured values can thus be available in vector form.

The spatial resolution of the first thermal image is generally clear, at least one or even two or three orders of magnitude coarser than the spatial resolution of the second thermal image. For example, the pixels of the first thermal image may have an edge length of at least about 1 km so that the area covered by a pixel is about 1 km ² or even more than 1 km ² or more than 3 km ² at the first thermal image. In contrast, the edge length of a pixel of the second heat image is preferably less than 100 m, or even less than 70 m or less than 50 m. Ideally, the edge length of a pixel of the second thermal image is in the range of only 10 m, so that the area imaged by a pixel of the second thermal image is only a few 100 m ² .

Alternatively, however, the method described here is also applicable if the spatial resolution of the first thermal image is less than an order of magnitude, for example only by a factor of 2 or 3, greater than the spatial resolution of the second thermal image. The method is applicable to any pair of images in which the spatial resolution of the second thermal image is higher, or less coarse, than the spatial resolution of the first thermal image.

In addition, the radiometric precision of the first thermal image is higher than the radiometric precision of the second thermal image. This feature of the thermal images is preferably achieved in that the Aufnahmevorrich tion of the first satellite via appropriate calibration technique, for example, by the use of black bodies or specially cooled th or temperature-controlled sensors has a high radiomet-metallic accuracy, preferably less than 2 K or, particularly preferably even under 1 K, ideally per pixel or even on average of the recorded thermal images guaranteed. To increase the predictability of the accuracy, the recording devices of the satellites can also provide data which, preferably per recorded pixel, indicates an accuracy of measurement estimated by sensors of the recording device. The radiometric precision of the second thermal image is lower, for example by the fact that the second thermal image has been recorded by the recording device of a low-cost small-sized satellite or another manned or unmanned aerial object which does not have a correspondingly precise calibration technique.

Furthermore, the present application comprises a method for determining the change of a measured variable within a region over a period of time, such as within an hour, a day or even within a week, a month or a year. By determining the temporal change of the measured variable of a territory as precisely as possible, for example, urban heat islands can be detected and their causes analyzed. Furthermore, such measures-time series provide knowledge for urban planning, for example the effect of green spaces, different roof types or areas of water on the urban climate.

The method described for determining the change in the measured variable within a region comprises first determining at least two, preferably a plurality of more than 10 or more than 100, as precise as possible thermal map of the area at different times, the two heat maps according to the above Procedures were determined.

In particular, the first thermal map was determined using first and second received thermal images of the area, wherein the first and second thermal images were taken synchronously or approximately synchronously at a first time and, of the first and second thermal images, the first thermal image was higher radiometric Precision and the second thermal image has a higher spatial resolution.

Further, the second thermal map was determined using third and fourth received thermal images of the area, the third and fourth thermal images being taken in synchronism or approximately synchronously at a second time and, of the third and fourth thermal images, the third thermal image being higher radiometric Precision and the fourth thermal image has a higher spatial resolution. From the first and second heat map of the area, a time series of the measured variable within the area can then be created.

Optionally, it is also possible in this way to determine heat maps of a plurality of times for an area and to generate a time series of the measured quantity within the area based on the plurality of heat maps of the area.

Preferably, the first and second times are within a bestimm th time period, for example within a day, or within a few hours or even within less than one or half an hour to ensure a high temporal resolution of the time series of the measurand within the area, for example to the Change in the measured quantity within one day.

In order to achieve such a high temporal resolution, it is possible, for example, that the first and the third thermal image are each taken by a large satellite platform with appropriate calibration technology to ensure high radiometric precision, such as a geostationä ren weather satellites. When using a first and third thermal image, which are recorded by geostationary satellites, preferably both the first and the third thermal image are taken up by the recording device of the same geostationary satellite. Alternatively, the first and third thermal images may also originate from different geostationary satellites with similar coverage.

Furthermore, it is alternatively also possible that the first and third thermal image are recorded by the recording devices of different large but not geosta tionary satellites in a low Erdorbit, the recording devices should be equipped with calibration technology for high measurement accuracy. If the first and third thermal images originate from different non-geostationary satellite platforms, then this is only possible if these two satellite platforms were within the predetermined time span over the area. Since non-geostationary large satellites, such as Landsats, have a temporal repetition rate of several ta with respect to a given area. gene until several weeks, such recordings within a given time window by different non-geostationary satellites th with currently available satellite infrastructure especially opportunistic possible, but unsuitable for creating continuous time series over a selected area with several, for example, more than 10 or more than 100 Image pairs, wherein the time intervals of two consecutively recorded image pairs each have a time interval of a few minutes or hours.

How many image pairs are necessary to create a time series, as well as the tolerable time intervals of the image pairs is generally strongly dependent on the application. For example, for some applications where temperature evolution is to be monitored over the year, one or two images per day may suffice, while for other applications where, for example, heat development during the day is to be investigated, a temporal resolution of only a few minutes is desirable is worth, for example, under 10 or under 30 minutes. To create a longer time series consisting of more than 10 or more than 100 image pairs, it is helpful if the first heat image with the higher radiometric precision was always or at least mostly taken by the same geostationary satellite of each synchronously recorded image pair.

In order to achieve a high spatial resolution at the same time, the second thermal image and the fourth thermal image are respectively recorded by recording devices of small satellites or other manned or unemann th flight objects that can provide a high spatial resolution of the recorded thermal images by their proximity to the earth.

These small satellites just described orbit the Earth generally in low Earth orbit. To realize a high temporal resolution, therefore, preferably the thermal images of several different small satellites, such as pico, nano, or microsatellites can be used, so that the second thermal image was taken by the receiving device of a first Kleinsatelli th while the fourth thermal image of the recording device a second small satellite was added. Since small satellites, such as CubeSats, are less expensive to produce compared to geostationary large satellites, it is possible to generate a time series with high temporal resolution of the measured quantity within a region.

The present description further includes a computer program product for calculating a thermal image, wherein the computer program product includes instructions that, when executed on a computer, perform the method described above. The instructions may optionally be instructions for determining a spatially highly resolved and simultaneously accurate thermal image of an area, or the instructions may be instructions for determining a time series of the measured variable within the area. In any case, the instructions are based on thermal images of different satellite platforms received by a corresponding receiving unit of the computer.

The computer can be a stationary PC or a mobile computer. The computer can also be part of a distributed system or a

cloud-based service suitable for executing program instructions.

The features described above with reference to the method or the device for creating a heat map can furthermore also be used in relation to the method, the device or the computer program product.

Further advantageous embodiments are described below with reference to the Figu ren. In this case, the temperature was chosen as the parameter because of its easy interpretability. Alternatively, however, the measurements may include other radiometric data such as radiances or other data acquired by the sensors of the satellites used or derived from the acquired measurements. Show it

Fig. 1 is a schematic representation of a device for determining a thermal map of an area;

Figure 2 is a schematic representation of a sequence of the method for Be mood of a heat map of an area. Fig. 3A Schematic representation of a recorded with a first satellite A thermal image of the earth's surface;

3B Schematic representation of a thermal image of the earth's surface recorded synchronously with FIG. 3B with a second satellite B;

Fig. 4A Schematic representation of a single pixel of the recorded by Satel lit A thermal image according to Fig. 3A;

Fig. 4B Schematic representation of the measurement of satellite B of the same region as shown in Fig. 4A;

4C Schematic representation of the particular heat map based on FIG. 4B and a specific temperature offset.

Figure 1 shows an apparatus for determining a heat map of a Ge territory. The device 101 comprises a receiving unit 102 which can receive signals from various satellites 104 and 105 which revolve around the earth 106. The reception of the signals of the satellites 104, 105 is usually indirect, that is not directly from the satellite, but via an intermediate device 107, which receive the signals of the satellite or satellites 104, 105 and then forward or make available for download. In particular, the receiving unit receives thermal imagery of the earth's surface taken by recording devices of the satellites 104 and 105. In order to produce a detailed thermal map of a high spatial resolution area in the range of a few meters to less than 10 meters and high radiometric accuracy, such as high temperature accuracy, the receiving device 102 first configures a first thermal image of a first satellite 104 and a second thermal image of a first second satellite 105 to receive.

Here, the first satellite 104 is a large satellite with high temperature accuracy, such as a geostationary weather satellite or a large earth satellite orbiting in a low earth orbit, such as a Landsat.

To achieve a temperature accuracy of preferably less than 2 K adds the first satellite 104 via built-in calibration technology. In the case of a geostationary first satellite, however, this first satellite 104 has only a very large spatial resolution capability because of its great distance from the earth, so that the pixels recorded have a size of several square kilometers. Alternatively, if a large satellite is used as the first satellite 104 in a low Earth orbit, such as a Landsat, the spatial resolution is already better than in the previously described geostationary satellites, but there is still room for improvement with respect to space higher resolution heat of a second satellite platform possible. For example, Landsat 8 currently has a resolution of 100 m. By using a small satellite as an auxiliary satellite, this resolution can be improved to well below 60 m. With post-processing of the data, even lower resolutions, for example below 30 m, are possible.

The second satellite 105 is a small satellite or CubeSat, which has a higher spatial resolution compared to the first satellite, but only one compared to the first satellite, reduced measurement accuracy.

The receiving unit 102 of the described device receives temporally and synchronously recorded first and second thermal images of spatially coregistered landscapes of the two satellites 104 and 105, that is, the recorded subregions of the earth's surface at least partially overlap. In order to now create a precise heat map of a region that is mapped on both thermal images, the device 101 further comprises a determination unit 103.

The determination unit 103 is configured to spatially allocate a pixel of the first thermal image of the first satellite to a group of pixels of the second thermal image of the second satellite and to calculate a measured quantity offset or temperature offset of the group of pixels of the second thermal image. This calculation is based on the assumption that the temperature deviation or measured variable deviation of the pixels of the second thermal image is at least locally constant, ie that the deviation of a pixel is not or only very slightly different from the deviation of adjacent pixels. Furthermore, it is assumed that the (measured with high precision) measured variable or temperature of the pixel of the first thermal image can be represented as a (weighted) sum or linear combination of the assigned pixels of the second thermal image. Thus, the measured variable offset for the group of pixels of the second thermal image can be determined from the relative measured variable differences of the pixels of the group of the second thermal image and the recorded measured variable of the assigned pixel of the first thermal image. The determination unit 103 is further configured to create a precise heat map of the area from the measured variable offset and the recorded measured variables of the second heat map.

FIG. 2 schematically shows a method for determining a precise heat map of a region, wherein the method can be implemented, for example, by the arrangement described above or by a computer program product which comprises corresponding instructions for carrying out the method.

The method described here for determining the precise heat map of a region comprises the following steps:

S201: receiving a first thermal image of a first region comprising transmitting a landscape captured by a first satellite receiver, the first thermal image comprising pixels spatially associated with the region and the thermal image respectively pixel-wise capturing the region one first Messwertzuweist.

S202: receiving a second thermal image of a second area encompassing landscape captured by a capture device of a second satellite, the second thermal image comprising pixels spatially associated with the region and the thermal image one pixel at a time assigns the second recorded reading.

It should be noted that steps S201 and S202 may be performed in any order or at the same time.

The received first and second thermal image must be used to carry out the Method additionally satisfy the following requirements: between the recording of the first satellite image and the recording of the second satellite image, there must be only a time offset below a fixed threshold, preferably less than 10 minutes or even less than 5 minutes, since the correlation of the thermal readings given in the two thermal images, with a greater time interval between the shots of the two images is not possible or only with difficulty. The time limit can be selected depending on the prevailing conditions. Under constant weather conditions, the longer distances between the two shots may be tolerated under circumstances as changing weather conditions.

Furthermore, a radiometric precision of the first thermal image, i. the accuracy of the recorded measured variable, higher than a radiometric precision of the second thermal image, whereas a spatial resolution of the second thermal image is higher than a spatial resolution of the first thermal image.

After receiving the two thermal images, the following further process steps can be carried out:

S203: determining a measurement offset of a first spatially associated pixel group of the second thermal image, wherein the first pixel group comprises a plurality of pixels by a weighted sum of relative measurements of the pixels of the first pixel group of the second thermal image versus the first acquired measurement the at least one pixel of the first thermal image, which is at least partially spatially associated with the first pixel group of the second thermal image. In this case, for example, the mean value of the measured values of the first pixel group is determined in order then to determine the offset, for example as a difference, to the assigned measured value or values of the first region of the first thermal image.

S204: determining corrected absolute measurements of the pixels of the first pixel group ordered to the region based on the second measured measured values of the pixels of the first pixel group and of the measured value offset. S205: Creation of the precise heat map of the area based on the corrected absolute measurements.

The above-described method for determining spatially high-resolution thermal maps with high measurement accuracy will be further described with reference to FIGS. 3A, 3B, 4A, 4B and 4C. The method presented here combines the radiometric accuracy of large satellite platforms with the spatial resolution of a second satellite platform, such as a small satellite or CubeSat. The method is based on the assumption that the specific intensity (radiance) of a pixel results from the linear combination of its subpixel components (linear mixing model). As a large satellite platform ideally one with a very high radiometric precision (Sentinel-3) and simultaneously high temporal resolution (Meteosat Second and Third Generation) is used.

The only prerequisite for the additional, smaller satellite is that it has the possibility of synchronous recording, so records IR images of the same region at the same time. The recorded images are shown schematically for both satellites in FIGS. 3A and 3B, in which case the temperature was selected as the relevant measured variable. In this case, satellite A (FIG. 3A) designates the platform with high radiometric precision and satellite B (FIG. 3B) the platform with high spatial resolution.

The drawn grid shows which areas of the satellite image B are to be assigned to the satellite of satellite A. The data from satellite B has an unknown temperature offset DG in the measurement of the radiation temperature, which does not allow an exact absolute temperature determination for satellite B alone. Relative temperature differences, for example between adjacent pixels, are correctly measured by satellite B (ensured by a previous calibration of the instrument).

By recording satellite A, the average of each subscribed te grid cell from the image of satellite B is known. Thus, the still unknown temperature offset DG can be excluded. This is given by the difference of the mean values of each cell of both images,

It applies

^ O ^ y) is the measured temperature of satellite A in the cell (x, y) and T _B (i, j) of the temperature measurement of satellite B in all pixels (i, /) inside this cell (x, y ) of satellite A, each with an areal fraction of w (i, j).

The measurement method is illustrated schematically for a single cell in FIGS. 4A, 4B and 4C: the mean value of all pixels of satellite B (FIG. 4B) shown does not yet correspond to the value of the corresponding pixel of satellite A (FIG. 4A). By subtracting a temperature offset resulting from the difference, the entire cell of the image is calibrated by satellite B and now has the same mean value as recorded by satellite A (Figure 4C).

There are several advantages over existing solutions:

The presented method combines the advantages of both measurements, high radiometric precision on the one hand and high spatial resolution on the other without having to access additional data sources. This enables resolutions of the order of a few meters, which is of great importance for applications in the field of UHI and satellite-based agriculture.

Due to the high spatial resolution of satellite B, no or very limited down-sampling is required and the number of surface classes per pixel is significantly reduced at high resolution, which increases the stability and accuracy of the surface temperature, taking into account land surface models and known emission levels the radiation temperature given by the particular heat map can be determined.

Furthermore, use of the unstable correlation between visual and IR images is avoided and the effect of temperature homogenization by choosing too few surface classes is significantly reduced. In addition, errors are avoided by a large resolution gap between ground resolution of the satellite sensor and resolution of the auxiliary data.

There are already thermal data with high radiometric precision, which can serve as a calibration standard for the concept. For example, Copernicus offers Sentinel-3 temperature data with <0.2 K accuracy at a resolution of 1 km ² . These data are freely accessible through the Copernicus program. Meteosat-9 offers about 9 - 15 km ² resolution with a temperature accuracy of about 2 K and a temporal resolution of a few minutes for large area coverage.

Moreover, with Meteosat-9 having a repetition rate of around 10-15, or up to 5 minutes across Europe, temporal and spatial synchronization of the two streams can easily be achieved.

Thus, the concept is also suitable for measurements with high temporal repetition rates when using multiple (small) satellites in a suitable constellation. Thus, a time series of the measured variable can be created within a region which enables an analysis and / or visualization of the change in the size of an area over a certain period of time, for example over the course of a day.

In this case, previously determined heat maps of a region are combined to form a time series, the recording times in each case being two consecutive heat maps of the time series each within a defined time span. As a recording time of a heat map in this case the time point of recording the heat map associated with this first and two th thermal images are considered or, with a small temporal Ver rate of the two synchronous recordings of the thermal images through the two satellites used, a mean time or a pair of both Recording times of the thermal images. The fixed period of time may vary depending on the application, but in many cases it is within a few hours, less than an hour, or even in the range of about 10 minutes. The inclusion of several consecutive images each within the specified period of time can therefore be done in many cases only by a geostationä ren satellite or by a plurality of satellites in a low Earth orbit.

The requirements for the IR detector of satellite B are relatively low, so that the approach is suitable for use in comparison to traditional satellite missions much cheaper small satellites or CubeSats. Thus, an improvement of the previously available spatial resolution by several orders of magnitude at a fraction of the cost of a traditional len satellite mission is possible. The maximum attainable spatial resolution is quasi limited only by the detector and the optics of the satellite.

A complex and complex absolute calibration of the detector aboard satellite B is eliminated, without thereby deteriorating the accuracy of the resulting surface temperature of the land. The solution does not require tracing or recognition of objects in the image and no ground measurements.

The significant increase in spatial resolution without the introduction of error sources generated by downsampling allows a variety of new or improved applications, for example in the localized prediction of traffic conditions (eg wet or traveled roads), in the environmental research (eg movement of gravel deposits on glaciers, Forest fires), in agriculture (vegetation health via evapotranspiration), in medicine (eg health risk of overheating) and, closely related to this, in urban planning (eg energy efficiency assessment of buildings or measurement of the urban microclimate).

The latter is particularly interesting as an application to investigate the effects of urban heat islands and corresponding countermeasures. Increasing the resolution of square kilometers to the size of a building or block of buildings provides important information on energy efficiency, temperature development, and the impact of green, green, or green spaces reflecting roofs or open water surfaces on urban climate. So far, corresponding surveys are based on a piecework of simulated models together with isolated soil measurements as well as data generated by one of the downsampling methods described above. A temporally and spatially high resolution method can help that

To plan cities of the future healthier and more environmentally friendly.

Claims

FRAUNHOFER-GESELLSCHAFT ... eV 197PCT 1297 Claims

A method of creating a thermal map of an area, comprising the following steps:

Receiving a first thermal image of a first region encompassing the landscape taken by a capture device of a first satellite (104), wherein the first thermal image comprises pixels spatially associated with the region and the thermal image provides the region pixel by pixel a first measurement value assigns (S201);

Receiving a second thermal image of a second region comprising the region taken by a host device of a second satellite (105), the second thermal image comprising pixels spatially associated with the region and the thermal image providing the region pixel-wise with a second one Assigns measured value (S202);

wherein between the recording of the first thermal image and the recording of the second thermal image, a time offset below half a predetermined barrier exists; and a radiometric precision of the first thermal image is higher than a radiometric precision of the second thermal image; and a spatial resolution of the second thermal image is higher than a spatial resolution of the first thermal image;

Determining a measured value offset of a first, spatially assigned, pixel group of the second thermal image, the first pixel group comprising a plurality of pixels, by comparing relative measurements of the pixels of the first pixel group of the second thermal image to the first measurement of the at least one pixel of the first Thermal image which is at least partially spatially associated with the first pixel group of the second thermal image (S203), Determining corrected absolute measurements of the pixels of the first pixel group ordered to the region based on the second measurements of the pixels of the first pixel group and the measured value offset (S204); and

Creation of the precise heat map of the area based on the corrected absolute measurements (S205).

2. The method of claim 1, wherein the method comprises, in particular: determining the measured value offset of the first pixel group of the second thermal image by comparing relative measured values of the first pixel group of the second thermal image to the first measured value of the spatially associated pixel of the first pixel group of the second thermal image of the first thermal image.

3. The method of claim 1, further comprising determining land surface temperatures of the area based on the precise heat map and land surface models and / or emission models.

4. The method according to any one of the preceding claims, wherein the thermal images are infrared images.

5. The method according to any one of the preceding claims, wherein the spatial resolution of the second thermal image is at least one or two orders of magnitude higher than the spatial resolution of the second thermal image th.

6. The method or arrangement according to one of the preceding claims, wherein the measured values are temperature values and a radiometric deviation of the first thermal image per pixel is below 2K and the radiometric deviation of the second thermal image in the center is greater than 2K.

7. The method according to any one of the preceding claims, wherein the first thermal image was recorded by a calibrated to increase the radiometric Präzisi on recording device of the first satellite men.

8. The method according to claim 1, wherein the first thermal image has been recorded by the recording device of a large satellite platform, for example a geostationary satellite in a high earth orbit, or a large, non-geostationary satellite in a low earth orbit.

9. The method according to any one of the preceding claims, wherein the second thermal image was taken by the receiving device of a Kleinsatelli th.

10. A method of determining a change in temperatures of an area over a period of time, comprising the following steps:

Determining a first thermal map of the area at a first time in accordance with the method of any one of claims 1 to 9 using first and second received thermal images;

Determining a second thermal map of the area at a second time according to the method of any one of claims 1 to 9 using a third and fourth received thermal image,

Creating a time series of the temperature of the area based on the first and second heat maps.

11. The method of claim 10, wherein the first and the second time point are within a predetermined period of time and the first and the third thermal image each have a high radiometric precision and were each taken from a first, large satellite platform, the second thermal image has high spatial resolution and was recorded by the recording device of a second satellite and the fourth thermal image has a high spatial resolution and was recorded by the recording device of a third satellite th, the second and third satellite are under different small satellites.

12. The method according to claim 10 or 11, further comprising the fol lowing steps:

Determining a plurality of additional heat maps of the area at additional times, each two consecutive points in time of the additional points being within a specified period of time, and

Creating a time series of the temperature of the area based on the first heat map, the second heat map and the additional heat maps.

A computer program product for determining a thermal map of a region or a method for determining a change in temperatures of a region, the computer program product comprising instructions that, when executed on a computer, perform the method of any one of claims 1 to 12.

14. Device (101) for creating a thermal map of a region comprising at least one receiving unit (102) configured for

Receiving a first thermal image of a first region encompassing the landscape taken by a capture device of a first satellite (104), wherein the first thermal image comprises pixels spatially associated with the region and the thermal image provides the region pixel by pixel a first measurement value assigns,

Receiving a second thermal image of a second region comprising the area taken by a host of a second satellite (105), the second thermal image comprising pixels spatially associated with the area and the second thermal image providing the area pixelwise with a second one Assigns metric wherein between the recording of the first thermal image and the recording of the second thermal image, a time offset below half a predetermined barrier exists; and a radiometric precision of the first thermal image is higher than a radiometric precision of the second thermal image; and a spatial resolution of the second thermal image is higher than a spatial resolution of the first thermal image;

Determination unit (103) configured for

Determining a measured value offset of a first, spatially assigned, pixel group of the second thermal image, wherein the first pixel group comprises a plurality of pixels, by comparing relative measured values of the pixels of the first pixel group to the first measured value of the at least one pixel of the first thermal image the first pixel group of the second thermal image is at least partially spatially associated;

Determining corrected absolute measurements of the pixels of the first pixel group based on the second measured temperature values of the pixels of the first pixel group and the measured value offset; Creating the heat map of the area based on the corrected absolute readings.